Approximation properties in \(L_ p\), \(0<p<1\) (Q1343493): Difference between revisions

Let \(f \in L_p(\mathbb{T})\) and \(\Lambda_n f = \sum^n_{\ell = - n} \theta ({{|\ell |} \over n}) \widehat f_\ell\) \(e^{i \ell x}\) where \(\theta \) is a \(C^\infty_0\)-function on \(\mathbb{R}\), such that \(\theta(x)=1\) for \(|x|\leq 1\) and \(\theta(x)=0\) for \(|x |> 2\). Theorem 1. If \(0 < p < 1\), then \[ |f - \Lambda_n f |^p_p \leq C (p)n^{p - 1} \min \left\{ \sum^n_{k = 1} \left( {E_k (f)_p \over k} \right)^p,\;\sum^\infty_{k = n + 1} \left( {E_k(f)_p \over k} \right)^p \right\}. \] Here \(E_k (f)_p\) denotes best approximation of \(f\) in \(L_p (\mathbb{T})\) by trigonometric polynomials of degree \(\leq k\). Let now \(f_n\) be a trigonometric polynomial of degree \(\leq n\) and \(f_n^{(\alpha)}\) denote its fractional derivative (in the Weyl sense) of order \(\alpha\). Theorem 2. If \(0 < p < 1\) and \(\alpha > 0\), then \[ \sup_{|f_n |_p \leq 1} |f_n^{(\alpha)} |^p_p \approx \begin{cases} n^{\alpha_p}, \quad & \text{if} \quad \alpha \in \mathbb{N} \text{ or } \alpha \notin \mathbb{N} \text{ and } \alpha p > 1 - p, \\ n^{1 - p}, \quad & \text{if} \quad \alpha p < 1 - p, \\ n^{1 - p} \log n, \quad & \text{if} \quad \alpha p = p - 1,\;\alpha \notin \mathbb{N}.\end{cases}. \]